Within LTE, the UEs can be in two modes: connected or free. In the first one, they introduce the handover process; in the second, they execute the initial cell selection and re-selection processes. The handover process allows to transfer a data session from one cell to another without interruptions. It can be intra-frequency, inter-frequency over the same system, or intersystem towards UMTS or GSM networks.

The handover is triggered via events (3GPP-TSGRAN, 2015) where either the serving or the neighbor cell exceed certain configured Reference Signal Received Power [RSRP] threshold. In LTE, the following intra Radio Access Technology [RAT] events exist:

• A1, the level of the serving cell is higher than the configured threshold;

• A2, the level of the serving cell is lower than the configured threshold;

• A3, the level of a neighbor cell is somewhat higher than the level of the serving cell;

• A4, the level of the neighbor cell is higher than the configured threshold; and

• A5, the level of the serving cell is lower than a first threshold and the level of the neighbor cell is higher than the second threshold.

In compliance with the previous events, two handover types are available: based on the A3 event, as shown in Figure 1; and based on the event A5, shown in Figure 2.

Figure 1
A3 event for the handover process

Figure 2
A5 event for the handover process

The HOM allows to configure the difference in the signal level of the target cell over the serving one, i.e., the RSRP level of the target cell shall be superior in that margin over the current cell. The TTT allows to delay the handover start time to avoid unnecessary handovers that produce a ping-pong effect. Consequently, the destination cell must maintain the highest signal level at least during the TTT to avoid the triggering of unnecessary handovers during variations of possible target cells.

The 3GPP specifies six alternatives regarding the architecture for mobile relays, called: Alt.1, Alt.2, eAlt.2-1, eAlt.2-2, eAlt.2-3, and Alt.4 (3GPP TSGRAN, 2014a). Within the Alt.1 architecture —also called Relay Full L3—, the SGW/PGW entities serving the relays network are separated from the donor cell. One of the advantages of this architecture is the fact of providing a simpler handover and during the switch of the return network link through the source and sink donor stations (means of transport in movement), a stable anchor point is maintained at IP level. This ensures the group mobility support of the users served by the relay (Krendzel, 2013).

In Figure 3, we show the handover process for the Alt.1 architecture of mobile relays, which is identical to the eNB handover process for a defined UE by the 3GPP (3GPP TSGRAN, 2017).

Figure 3
Handover process for the Alt.1 of 3GPP TR 36.836 mobile relays (3GPP-TSGRAN, 2010)

Here, we present the works involving modifications to the standard network architecture or proposing a different handover type than the hard handover used in LTE.

Wang, Ren, and Tu (2011) describe a soft handover for TD-LTE in a high-speed trains scenario by seeking the reduction in the number of Radio Link Failure [RLF] and the ping-pong effect. Furthermore, the proposal pursues the minimization in the interruption time caused by the high speeds to provide a satisfactory user experience. The algorithm grants a list for each UE inside the scenario and each of these handover lists has the control information in the downlink of two neighbor eNBs. Nevertheless, only a single transmission in the downlink is considered from the eNB with better RSRP. The algorithm is based on the eNodeB adding/removal steps to the list, depending on the RSRP thresholds.

On the other hand, the semi-soft handover algorithm for multi-bearer systems based on the Site Selection Diversity Transmission [SSDT] is a macro-diversity method employed in soft handovers over WCDMA networks. Its objective is to reduce the inter-cell interference caused by multiple transmissions while the transmission is performed in the downlink from the best serving cell. In the semi-soft handover, the bandwidth is divided in two different bands: control and data. The control band, with a frequency reuse factor of 7, is divided in 7 sub-bands, each one of them handled by a cell. A UE can —simultaneously— receive information from all the neighbor cells through the control channels and select the cell with higher RSRP to transmit packets via SSDT. According to the simulations performed in the study, a better performance in the number of successful handovers is obtained; nevertheless, when control signals are received from all the neighbor cells, the number of headers in the signaling is increased (Lee, Son & Lee, 2009).

The scheme proposed by Lee, Chuang, Chen, and Sun (2014) is based on LTE femtocells using multiple output interfaces —called Multiple Egress Network – Network Mobility [MEN-NEMO]— to reduce the latency and the number of headers in the handover process. For the network core architecture, the authors propose the usage of improved femtocells deployed in the train to provide service to the passengers’ UEs. In accordance with the performed simulations, the proposed scheme was able to reduce the latency and the number of headers in the signaling for the handover processes.

Here, we present some algorithms that use the UE speed and position to trigger the handover process. Luan, Wu, Shen, Ye, and He (2012) propose two types of LTE handover algorithms for high speed trains, called fast handover algorithms: one based on the UE speed and the other based on the TTT optimization. The first one proposes to adjust the handover hysteresis and TTT dynamically and according to the train speed to ensure a successful handover. With this, when the speed is increased, both the hysteresis and the TTT suffer small decrements. According with the simulations, the Handover Success Rate [HSR] is improved for certain values of Signal to Interference plus Noise Ratio [SINR]. Within one second, the RSRP and RSRQ variations from the serving and destination cells produced by the train speed are considered. Therefore, the authors propose to adjunt the hysteresis margins and, to avoid the ping-pong effect, they define a number of times N that the target cell measurements exceed the source ones. With these conditions, the TTT will dynamically vary to reflect the channel conditions and it will avoid handover failures and the ping-pong effect. Similar to the previous case, given the simulation results, the HSR is improved for certain SINR values.

The work presented by Zhang, Wu, Zhang, and Luan (2014) analyzes a high-speed trains scenario, where the train direction is considered together with the GPS data, which are relatively easy to obtain. The neighbor cells through the railways are also known; hence, they propose to use this information to implement a fast handover scheme. Their proposal settles appropriate reference points to trigger a handover procedure and their algorithm allows to reduce the latency and having smaller overlapping areas for the handover. The selection of an adequate reference point (geographic location) to perform the handover is fundamental; consequently, the points correspond to the statistical result of repetitive measurements. With the use of the hysteresis parameter and the measurement of the Block Error Rate [BLER], the authors tried to determine the geographical points where the handover must be triggered to send the session towards the destination cell. As per their simulation results, the ping-pong effect is reduced and they obtained a favorable radiolink failure rate.

We present some algorithms using an adjustment in the handover parameters based on the received signal levels. The work of Anas, Calabrese, Mogensen, Rosa, and Pedersen (2007) uses a TTT window based in the Received Signal Strenght [RSS]. The instantaneous RSRP value is stored in a record, whose values are later processed to determine if the handover must be triggered. The algorithm is implemented in two stages: storage and comparison in the signal level. As per their simulation results, they suggest HOM values to obtain a good relation between the number of executed handovers and the SINR.

The algorithm proposed by Zheng and Wigard (2008) considers the handover decision in the usage of historical records relative to differences in the signal level. The algorithm is executed in three stages: the calculation of the RSRP differences, the processing of the filtered RSRP values, and the handover decision. The simulation results show a reduction in the number of executed handovers by the UE; nevertheless, this algorithm does not consider the TTT, since the ping-pong effect is present.

The Hard Handover Algorithm with Average Received Signal Reference Power Constraint [LHHAARC] is a proposal seeking to minimize the number of handovers and delays, maximizing the data transfer in the network. LHHAARC is based on the hard handover algorithm with an additional condition over the average RSRP to improve the efficiency when the handover is executed (Link, Sandrasegaran, Ramli, & Basukala, 2011).

The concept of the previous algorithm is mainly to limit the handover possibilities to minimize the unnecessary executions of it, ensuring the channel quality in the target cell by having a higher RSRP than the serving cell. This is executed by having certain threshold from the first to the last measurement period. Their results indicate that the number of handovers is reduced compared with the TTT window algorithms based on RSS and integrator. Furthermore, lower delay values were obtained, having a higher data transfer rate.

In this section, we describe research works showing algorithms operating with mobile femtocells or mobile relays. Sui,Vihriälä, Papadogiannis, Sternad, Yang, and Svensson (2013) study the deployment of the so-called Moving Relay Nodes [MRN] in public transportation vehicles to provide service to the passengers within them. They analyzed the behavior relative to the performance of a fixed relay. For the study, several simulations were carried out using a fixed relay at different distances from a vehicle —where its position is assumed as known—. Their results show that the users inside the vehicles are affected by the so-called Vehicular Penetration Loss [VPL] and that their algorithm considerably improves the user data transfer by reducing those VPL. This was achieved by locating the relay as close as possible to the vehicle, where the reader can infer that the relay should move with the vehicle, entailing in the concept of mobile relay.

Karimi et al., (2012) present a solution based on LTE networks to support a high data transfer and multimedia services in high speed trains. The solution employs an array of organized cells throughout the railways, together with femtocells supporting the traffic demand inside the trains. These last are called mobile eNB or mobile femtocells. Additionally, they use a handover process called predictive, where the cells array allows to know in advance the target cells. Their simulation results show that improvements at the transfer level are achieved, together with a low handover latency and a higher success rate in the handover execution.

Chen (2015) researches the benefits of the mobile relays in a scenario where they are deployed in public buses. Although the author does not specifically propose a handover algorithm, his work allows to observe the benefits of the mobile relays. His research is focused on three aspects: improvements in the mobility procedures for mobile relays, where several signaling procedures are studied and the same protocol stack of the fixed relays is employed; interference impact in the transfer rate of a vehicular UE, where the transfer rate of a UE connected to the macro station is compared with the one of a UE connected with the mobile relay; and impact of the mobile relay in the cell capacity, where the capacity gain obtained by the relay deployment is assessed in a scenario with multiple users and relays.

Davaasambuu and Sato (2014) propose an adaptive hysteresis scheme based on a cost function, which considers some relevant factors of the mobile relay such as the cell load, relay speed, and the service type requested by the UE. The scheme operates between handover donor cells via the X2 interface. Their results indicate that the proposed scheme presents a higher probability of a successful handover than a scheme based on SINR.

Davaasambuu, Yu, and Sato (2015) present a self-optimization hysteresis scheme with mobile relay dual nodes in high speed environments. They propose an adjustment mechansim of the handover parameters such as individual cell histeresis and offset based on the vehicle speed. Further, they propose a Handover Performance Indicator [HPI] as a measurement for their scheme. This HPI helps to monitor the handover performance and it is equal to the sum of the failure handover rate, ping-pong handover, and failed handovers due to radio link failures. The simulation results show that, when comparing the proposed scheme with the traditional one, the authors’ one can reduce the number of radio link failures and service interruptions during the handover process.

On the other hand, Zhao, Huang, Zhang, and Fang (2011) propose an algorithm for the handover decision based on the Relative Velocity Aided Handoff [RVAH] of the UE towards the access points (cell, fixed or mobile relay), server, and destination. If the UE presents speeds lower than a threshold V0 regarding the server and destination access points, the handover will be performed according to the HOM and TTT; that is, the standard way. If the speed regarding the server access point is higher than the threshold V0, the handover process is triggered as per the RVAH algorithm. The neighbor access points with relative speeds lower than V0can be selected as future destination points.

Davaasambuu, Semaganga, and Sato (2015) propose an adaptive hysteresis scheme (based on the train speed) for handover in wireless networks with mobile relays in a high-speed trains scenario. They include a logarithmic formula involving the maximum and current speed of the mobile relay; also, they involve hysteresis values called “by default” and “maximum” for the hysteresis adjustment in the handover process. The probability of interruptions in calls during the handover is reduced by introducing a modified call admission control to support the savings in radio resources. This helps to prioritize the handover of the calls of the mobile relay over other calls. The authors present three priorities: mobile relay handover calls, UE handover calls, and new calls. The results described in their paper show how the successful handovers rate is increased and the radio failures is reduced.